Strategic decision points in experiments: A predictive Bayesian optional stopping method

Abstract

Sample size determination is crucial in experimental design, especially in traffic and transport research. Frequentist statistics require a fixed sample size determined by power analysis, which cannot be adjusted once the experiment starts. Bayesian sample size determination, with proper priors, offers an alternative. Bayesian optional stopping (BOS) allows experiments to stop when statistical targets are met. We introduce predictive Bayesian optional stopping (pBOS), combining BOS with Bayesian rehearsal simulations to predict future data and stop experiments if targets are unlikely to be met within resource constraints. We identified and corrected a bias in predictions using multiple linear regression. pBOS shows up to 118% better cost benefit than traditional BOS and is more efficient than frequentist methods. pBOS allows researchers to, under certain conditions, stop experiments when resources are insufficient or when enough data is collected, optimizing resource use and cost savings.